Meta 推出 Code Llama：基於 Llama 2 的開源程式碼生成模型，支援 100k 上下文
Original: Code Llama: Llama 2 learns to code
Meta has officially launched Code Llama, a family of state-of-the-art open-source code generation models fine-tuned on Llama 2. Code Llama…
Meta 正式發布專為程式碼任務設計的開源模型 Code Llama，提供 7B、13B 與 34B 三種尺寸，並包含 Python 專用版與指令微調版。該模型支援高達 100k 的上下文視窗，在多項程式碼基準測試中表現優異。Hugging Face 已同步支援該系列模型的推理、部署與微調。
Meta has officially launched Code Llama, a family of state-of-the-art open-source code generation models fine-tuned on Llama 2. Code Llama achieves leading performance among open-source models across multiple coding benchmarks, even approaching GPT-3.5 on certain tests.
Free shows the 3-line summary; Pro unlocks the full deep summary (~300 words) so you never have to click through.
See Pro plans →Want the original English / full article?
Read on Hugging Face Blog →Summaries are AI-generated; the original article is authoritative.